Threadmark

About The App

Threadmark is a private attention inbox for the WhatsApp groups and people you choose. It reads new messages as they arrive, picks out payments, meetings, deadlines, and anything else you ask it to watch for, and shows only those. Everything else stays in WhatsApp.

It connects as a linked device, like WhatsApp Web, and is read-only by design: it never sends messages, never marks chats as read, never shows you as online, and never reads old history. The connection is unofficial, so WhatsApp could change it or restrict an account that uses it.

Every group and person is off until you switch them on. Messages from other chats are discarded before they are checked or stored, and selected messages that match nothing are not kept.

Gallery

The chats in these screenshots are made up.

How it works

1. Connect WhatsApp. Link Threadmark as a device with a QR code or a pairing code.
2. Choose the chats. Turn on only the groups and people to watch. Search by name or number, or import contacts from your phone.
3. Let it sort. Built-in detectors and your own phrase rules check each new message on the server. If nothing matches and Jev is on, the AI service takes a second look.
4. Act on what matters. Mark items done, snooze them, add them to a calendar, or tell Threadmark when it got something wrong.

Features

• Payments and meetings, found locally: Built-in detectors read Romanian and English messages and pull out amounts, currencies, IBANs, dates, times, and links.
• Your own rules: Phrase rules match any or all of your words, ignoring case and accents, and never leave the server. Each one can be limited to some chats, send an alert or not, and be tested on a sample first.
• Monitors in plain language: Describe what to watch for in a sentence, such as a parcel or a reservation that needs attention, and choose how sure Jev must be before it flags a message.
• AI only as a fallback: When nothing matches locally, the optional Jev service from TypeSafe AI checks for payments, meetings, and reminders and scores urgency. It receives only the message text, never the chat name, the sender, phone numbers, or WhatsApp IDs.
• Deadlines and urgency: Due dates are read from the message, items show due and overdue badges, and urgent ones rise to the top.
• Payment warnings: Changed bank details, changed amounts, possible duplicates, and unusual requests for passwords or payments are flagged for a second look.
• Attachments, read locally: Photos, PDFs, and voice notes can be read with text recognition and speech transcription on the server, then deleted straight away. A handwritten note it cannot read becomes a reminder to check the photo.
• Changes and replies: Threadmark can notice when a plan is corrected or cancelled, and close a reply reminder once you answer. Both are off by default and keep at most 48 hours of recent text.
• Alerts and a daily digest: Push notifications for new items, an urgent title for urgent ones, and an optional daily summary at the time you choose.
• Kept short: Matched excerpts are deleted after 30 days by default, unless their item is still open.

Under the hood

Threadmark has two parts, each in its own rootless Podman container. A bridge holds the WhatsApp connection through the Baileys library and keeps every message in a queue until the app has received it. The app runs on Node.js 24 with its built-in SQLite and sends live updates to the browser through server-sent events and Web Push. The bridge opens no port to the outside. Access is by invitation, managed from PWA Invite Console, and the app updates itself with pwa-kit. Messages reach the app in a neutral format, so other sources, such as Telegram or email, could be added later. The project is open source under the GPL-3.0 licence.